Deep Reinforcement Learning-Based Optimization for Energy Efficiency and Latency Reduction in UAV–WSN Systems

Abstract

The integration of Unmanned Aerial Vehicles (UAVs) with Wireless Sensor Networks (WSNs) offers promising potential for intelligent environmental monitoring and data transmission. Optimizing energy consumption and minimizing latency remain key challenges in dynamic communication environments. This work introduces a Deep Reinforcement Learning-Based Optimization Framework that enhances energy efficiency and reduces latency through adaptive routing, mobility control, and intelligent resource allocation. A Hybrid UAV–WSN Benchmark Dataset (HUW-BD) was utilized, combining simulated and real-world telemetry, transmission, and network state data. The proposed model integrates a convolutional feature extractor with an actor–critic learning policy for continuous decision-making. Comparative results demonstrated superior performance over LEACH, Q-learning, and ACO-based algorithms, achieving 98.9% accuracy, 0.95 J/packet energy efficiency, and 92 ms latency, while maintaining 85 Mbps throughput and 14 ms computation overhead. Experimental findings highlight the model’s effectiveness in balancing network stability, reliability, and energy-aware communication. This approach establishes a robust pathway for sustainable UAV–WSN systems in next-generation intelligent networking applications.